China's Moonshot AI released Kimi K3, described as the largest open-weight AI model publicly available.
The model has about 2.8 trillion parameters and a context window of one million tokens.
Its weights were made available for free public download, so developers can self-host and adapt it.
It is positioned as an open alternative to leading proprietary models from US firms.
The release reflects China's push to advance AI despite US curbs on advanced computing chips.
An AI language model's 'parameters' are the internal values it learns during training — more parameters generally mean greater capacity. Its 'context window' is how much text it can consider at once, so a one-million-token window lets it read entire large documents or codebases in a single query. An 'open-weight' model is one whose trained parameters are released publicly, so anyone can download, run and adapt it themselves — in contrast to proprietary models like the leading US systems, which are used only through their owners' services.
Simple Analogy: Parameters are the model's learned 'knowledge dials', the context window is how much it can hold in mind at once, and open-weight means the recipe is handed out rather than kept locked.
GS Paper 3 > Science and Technology; Emerging Technologies
General Awareness > Science and Technology Current Affairs
An AI model whose trained parameters are released publicly for anyone to download and run.
An internal value a model learns in training; parameter count is a rough measure of capacity.
The amount of text (in tokens) a model can consider in a single query.